RESUMO
The scarcity of high-quality data presents a major challenge to the prediction of material properties using machine learning (ML) models. Obtaining material property data from experiments is economically cost-prohibitive, if not impossible. In this work, we address this challenge by generating an extensive material property dataset comprising thousands of data points pertaining to the elastic properties of Fe-C alloys. The data were generated using molecular dynamic (MD) calculations utilizing reference-free Modified embedded atom method (RF-MEAM) interatomic potential. This potential was developed by fitting atomic structure-dependent energies, forces, and stress tensors evaluated at ground state and finite temperatures using ab-initio. Various ML algorithms were subsequently trained and deployed to predict elastic properties. In addition to individual algorithms, super learner (SL), an ensemble ML technique, was incorporated to refine predictions further. The input parameters comprised the alloy's composition, crystal structure, interstitial sites, lattice parameters, and temperature. The target properties were the bulk modulus and shear modulus. Two distinct prediction approaches were undertaken: employing individual models for each property prediction and simultaneously predicting both properties using a single integrated model, enabling a comparative analysis. The efficiency of these models was assessed through rigorous evaluation using a range of accuracy metrics. This work showcases the synergistic power of MD simulations and ML techniques for accelerating the prediction of elastic properties in alloys.
RESUMO
One of the major impediments to the computational investigation and design of complex alloys such as steel is the lack of effective and versatile interatomic potentials to perform large-scale calculations. In this study, we developed an RF-MEAM potential for the iron-carbon (Fe-C) system to predict the elastic properties at elevated temperatures. Several potentials were produced by fitting potential parameters to the various datasets containing forces, energies, and stress tensor data generated using density functional theory (DFT) calculations. The potentials were then evaluated using a two-step filter process. In the first step, the optimized RSME error function of the potential fitting code, MEAMfit, was used as the selection criterion. In the second step, molecular dynamics (MD) calculations were employed to calculate ground-state elastic properties of structures present in the training set of the data fitting process. The calculated single crystal and poly-crystalline elastic constants for various Fe-C structures were compared with the DFT and experimental results. The resulting best potential accurately predicted the ground state elastic properties of B1, cementite, and orthorhombic-Fe7C3 (O-Fe7C3), and also calculated the phonon spectra in good agreement with the DFT-calculated ones for cementite and O-Fe7C3. Furthermore, the potential was used to successfully predict the elastic properties of interstitial Fe-C alloys (FeC-0.2% and FeC-0.4%) and O-Fe7C3 at elevated temperatures. The results were in good agreement with the published literature. The successful prediction of elevated temperature properties of structures not included in data fitting validated the potential's ability to model elevated-temperature elastic properties.
